Anisotropy of magnetic susceptibility (AMS) data analysis is a convenient method used to investigate strain and flow during lava flow emplacement. In order to make a sound interpretation, the origin of the AMS signal must be verified. Two questions must be answered: 1) what phase, or phases carry the AMS signal and 2) when was the AMS fabric acquired? The verification steps themselves can provide extra data for interpreting lava flow conditions. Here, we present a methodology to answer the two questions in a 6 km-long Chaîne des Puys trachybasaltic lava flow that descended into the future site of Clermont Ferrand (France) 45,000 years ago. Knowledge of lava flow emplacement will be useful specifically to this site, if a reactivation of the volcanic chain occurs. The results are also of more general interest to understand lava flow emplacement dynamics. Curie Temperature, thin sections (optical, SEM, microprobe analysis) and First Order Reversal Curves (FORCS) indicate that the AMS carriers are multidomain (MD) and pseudosingle-domain (PSD) titanomagnetites. 3-D microlite fabric data compared with AMS-fabric elements shows that the AMS ellipsoid is produced by a late-stage microlite fabric deformation, just before complete immobilisation. With this knowledge, and with field structural observations, a vertical section through the lava flow is analysed. Two AMS parameters: magnetic lineation (kmax), and degree of anisotropy (A) are significant. The kmax displays opposing plunge directions, suggesting reversing simple shear sense. Some kmax plunge reversals coincide with degree of anisotropy breaks, also indicating the existence of texturally distinct units. Also, kmax plunges can be greater than 45° indicating a vertical pure shear component consistent with inflation. Degree of anisotropy breaks and kmax changes correlate with variations in vesicle and/or phenocryst concentrations, underlining that the distinct layers could have been rheologically different. Based on these observations, we propose a qualitative late-stage velocity profile for the flow that requires several distinct layers. We propose that the flow was layered into distinct rheology units, linked to pre-eruption or in-flow variability. This suggests that during the late stage of emplacement, the flow was subdivided into at least 5 distinct compartments, and each had a different flow history and different behaviour.